Automated tuning of analog neuromimetic integrated circuits

Neuromorphic engineering often faces the adjusting of the neuromimetic systems. Indeed, adjusting the parameters of integrated circuits and systems is a shared issue to address for the designers of tunable systems. This paper presents an original method to automatically tune reconfigurable neuromimetic analog integrated circuits according to biological relevance. This method is based on an evolutionary optimization technique, the Differential Evolution (DE) algorithm that had never been used for biological neuron modeling. To illustrate the adjusting method, we show how to reproduce the behavior of two kinds of well-known neurons, inhibitory and excitatory, by an automated tuning of the parameters of neuromimetic circuits. The behavior of the hardware neurons is then compared to the model one.